Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • A question from an absolute beginner: hospitalization data

    Hi everybody.
    I have a set of hospital discharge records (time window 2001-2016) with the main diagnosis for a cohort made up by about 5,000 former petrochemical workers. It is a follow up of the cohort studied years ago with, of course, a shorter time window (2001-2006).
    In the previous occasion, Hospital Discharge Odds Ratios (HDORs) by job category (reference White collars) were calculated.
    Does any of forum participants have a suggestion for the analysis to perform now?
    I am wondering whether to use the same approach, or to widen it, mainly for the length of the period.
    Thank's a lot for help.

  • #2
    Maybe you should think about IRR using "los" (length of stay) as the outcome variable. If so, a Poisson-like model (Poisson or Negative binomial regression) could be an interesting approach. If los starts by 1 (I mean, no zeroes), you can test a negative-truncated version of both models.

    Hopefully that heps.
    Best regards,

    Marcos

    Comment


    • #3
      Originally posted by Marcos Almeida View Post
      Maybe you should think about IRR using "los" (length of stay) as the outcome variable. If so, a Poisson-like model (Poisson or Negative binomial regression) could be an interesting approach. If los starts by 1 (I mean, no zeroes), you can test a negative-truncated version of both models.

      Hopefully that heps.
      Dear Marcos,
      thanks a lot for your message.
      My post was not very clear, I didn't notice this while writing it.
      What I meant was how to study the occurred diseases taking into account the length of residence in the studied town. This is because I am evaluating a seriously polluted area, and the underlying hypothesis is that the more you lived there, the more you can suffer from environmental pollution related diseases.
      Sorry for my obscure text.

      Best regards,

      Amerigo

      Comment


      • #4
        Well, if I understood right, generally speaking, you may have the number of environmental pollution related diseases as the (count) DV and a Poisson-like model would still fit in the scheme. Adjustment can be made by using the sector population as the "exposure" variable.
        Best regards,

        Marcos

        Comment


        • #5
          Dear Marcos,
          thanks for your good advice, and sorry for keeping you waiting for my answer.
          Best regards,
          Amerigo

          Comment

          Working...
          X